In the current state of the art of remote sensors, data are separately acquired from each of the remote sensors and separately stored in a storage device. The data may then be processed and analyzed to improve recognition of the remote target.
Recognition of a target without prior knowledge, however, is similar to searching and recognizing a needle in a haystack. Recognition of a target based on data collected from multiple sensors is very costly and inefficient. Furthermore, robustness of target recognition based on current fusion of multi-sensor data is questionable.
A typical target of interest covers an area that is likely smaller than 5% of the total area searched by a sensor. As such, much of the data collected and stored in a storage device is useless and, typically, is discarded after the data is retrieved from the storage device, processed and analyzed.
As will be described, the present invention provides an efficient, real-time and robust recognition of a target, based on real-time multi-sensor image acquisition and fusion of the acquired data that only pertains to the target region of interest.